Scholarlink

<p>During the production of corn straw feed, materials are prone to forming agglomerates in the screw conveying process, which significantly affects conveying stability and efficiency. Real-time detection of corn straw agglomeration can offer guidance for adjusting equipment operating conditions, ensuring stability throughout the conveying process. To this end, a real-time detection algorithm named RTAD-YOLO is proposed, based on YOLO11n. Firstly, the backbone network is replaced with a novel Shuffled FasterNet backbone, which combines PConv with a channel shuffle mechanism to achieve powerful yet lightweight feature extraction. Secondly, the integration of the BRA mechanism into the C2PSA module resulted in the C2BRA module, which enhances the algorithm’s ability to extract small-scale features through a bilevel routing strategy while reducing network latency. Finally, the neck network incorporates the C3k2-AD module, which integrates the efficient AdditiveBlock into the original C3k2 module, achieving a seamless combination of local feature extraction and global context modeling, thereby compensating for potential performance losses due to the lightweight backbone network. Experimental results on the self-constructed agglomerate dataset show that RTAD-YOLO achieved a mAP50 of 87.8%, surpassing YOLO11n by 1.3%, with P improving by 7.1%, R by 0.2%, parameter count decreasing by 7.7%, GFLOPs by 9.5%, and FPS increasing by 8.3%. Comparing to other mainstream algorithms, RTAD-YOLO achieves the optimal balance between detection accuracy and speed, providing robust technical support for real-time detection of corn straw agglomerates in screw conveying.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

RTAD-YOLO: a real-time agglomerate detection algorithm for corn straw in screw conveying

  • 期刊论文

  • 发表日期：

    2025年12月5日

查看全文
  • Wenbin Guo,
  • Xuhui Zhang,
  • Wulantuya

摘要

During the production of corn straw feed, materials are prone to forming agglomerates in the screw conveying process, which significantly affects conveying stability and efficiency. Real-time detection of corn straw agglomeration can offer guidance for adjusting equipment operating conditions, ensuring stability throughout the conveying process. To this end, a real-time detection algorithm named RTAD-YOLO is proposed, based on YOLO11n. Firstly, the backbone network is replaced with a novel Shuffled FasterNet backbone, which combines PConv with a channel shuffle mechanism to achieve powerful yet lightweight feature extraction. Secondly, the integration of the BRA mechanism into the C2PSA module resulted in the C2BRA module, which enhances the algorithm’s ability to extract small-scale features through a bilevel routing strategy while reducing network latency. Finally, the neck network incorporates the C3k2-AD module, which integrates the efficient AdditiveBlock into the original C3k2 module, achieving a seamless combination of local feature extraction and global context modeling, thereby compensating for potential performance losses due to the lightweight backbone network. Experimental results on the self-constructed agglomerate dataset show that RTAD-YOLO achieved a mAP50 of 87.8%, surpassing YOLO11n by 1.3%, with P improving by 7.1%, R by 0.2%, parameter count decreasing by 7.7%, GFLOPs by 9.5%, and FPS increasing by 8.3%. Comparing to other mainstream algorithms, RTAD-YOLO achieves the optimal balance between detection accuracy and speed, providing robust technical support for real-time detection of corn straw agglomerates in screw conveying.